The paradigm of fear conditioning remains a dominant model for characterizing the neurobiology of posttraumatic stress disorder (PTSD) (
1). Previously neutral stimuli that were bystanders in the setting of an aversive stimulus (e.g., trauma exposure) subsequently trigger fear and anxiety. Healthy adaptive responses to trauma involve successful extinction learning, whereby conditioned stimuli regain their neutrality. Extinction learning is a tractable process with excellent translation across mammalian species, and a body of basic science has implicated the role of glutamatergic signaling (
2). The
d-cycloserine story is one of the few examples of the development of novel pharmacotherapy emerging from neuroscience and not from serendipity. Investigators at Emory University led by Michael Davis, Barbara Rothbaum, and Kerry Ressler initiated a series of elegant studies demonstrating that
d-cycloserine, a partial
N-methyl-
d-aspartate (NMDA) agonist, can accelerate extinction learning in fear-conditioned animals (
3,
4) and in simple phobia, social anxiety disorder, and obsessive-compulsive disorder in human subjects (
5–
8). Since then, the question of whether
d-cycloserine can enhance exposure therapy for PTSD has been tested in a number of small trials, and results to date have not been consistent (
9–
11).
In this issue, Rothbaum and colleagues report the largest ever (N=156) randomized controlled trial of
d-cycloserine augmentation of exposure therapy in any population with PTSD (
12). Most notable is the fact that they tested the treatment in Iraq and Afghanistan war veterans with chronic PTSD—a population in which few controlled treatment trials have been conducted and which is potentially more difficult to treat than civilians with single-event trauma. Also notable is that the investigators included an active comparator, alprazolam, which potentially can interfere with extinction learning (
13). All participants were treated in six sessions with exposure therapy using virtual reality technology. The primary outcome measures involved rater- and self-report PTSD symptom scales, and a key secondary outcome assessed was a measure of extinction learning that was defined by the mean decrease in subjective units of discomfort (using a 0–100 scale) across exposure sessions (a large drop in distress across exposure sessions implies successful extinction). Rater-based and self-report measures were recorded at baseline, after six sessions of treatment, and at 3-, 6-, and 12-month follow-up assessments. Additional secondary outcome measures included two objective biomarkers: cortisol response to 2-minute virtual reality scenes, and startle response (orbicularis oculi electromyography) during scene presentation, obtained before and after the six-session trial period.
The results from this trial are a major contribution, and a newsworthy one, by virtue of the impressive sample size, the sample population, the rigor of the design, and the execution of the trial. The results showed that all three groups exhibited a substantial improvement in symptoms over the course of the trial, although, as the investigators were careful to explain, there was no treatment control condition. Overall, there was no treatment-by-condition interaction over time for either d-cycloserine or alprazolam against placebo over the full course of the trial and follow-up period. Rater-based PTSD symptom severity was higher in the alprazolam group compared with placebo at one time point after the trial.
A compelling aspect of this trial pertains to the measures of extinction learning—cortisol level and startle reactivity. The results demonstrate that changes in their measure of extinction learning favorably influenced the changes in the rater-based and self-report PTSD outcome measures only in the group receiving
d-cycloserine. Furthermore, pre- to posttreatment decreases in cortisol and startle reactivity were observed only in the
d-cycloserine group. Although there was no overall treatment advantage for
d-cycloserine, there was a noticeable favorable effect on objective biomarkers. Thus, this trial has elements that will appeal to everyone and may function as a projective test for one’s beliefs about the centrality of extinction in PTSD treatment or the validity of currently available biomarkers. If you’re a skeptic, you might focus on the negative overall results and dismiss the significance of the secondary biomarkers. If you have an affinity for silver linings, the fact that
d-cycloserine showed promise in measures of extinction learning and startle reactivity is an exciting realization of the promise of animal-human translational science. The trial results may also serve as a Rorschach test for your beliefs about benzodiazepines. If you believe they are toxic, then the one time point at which alprazolam showed a disadvantage compared with placebo will shine brightly against the background of the lack of difference with placebo across the full trial. If you are concerned that the forcefulness of existing treatment guidelines (
14) for benzodiazepine use in PTSD is not commensurate with the strength of the empirical evidence—either positive or negative—then this trial will support your view that clinical recommendations are supported by weak evidence.
The trial also has a number of puzzling results that can only be resolved by future trials. One question is why it is that an intervention that appears to have succeeded in reducing reactivity to trauma-related cues, presumably via extinction learning, did not produce an overall treatment advantage. Either the effects are not strong enough for d-cycloserine specifically, and other agents should be developed with better potency for facilitating extinction, or therapeutic improvement, even in the context of a seemingly pure exposure modality, involves more than extinction learning. Furthermore, given that exposure was the key element of treatment in all three groups, it is surprising that between-session extinction learning did not correspond to clinical gains in all groups. The cortisol reactivity is interesting, but it is hard to understand why the d-cycloserine group showed a posttreatment drop in cortisol in the context of the exposure challenge. The results do not demonstrate that exposure to the virtual scenes elicited a strong posttreatment cortisol response in any of the groups.
The results from this trial do not invalidate the strategy of pharmacologic augmentation of exposure therapy. This approach continues to have great promise, particularly with simple phobias and social anxiety disorder. However, the trial data from this report suggest that agents that enhance extinction learning will need to have greater potency than d-cycloserine to make a significant mark on combat-related PTSD.